Back
AppleNovember 19, 20262 sources

Apple building an autonomous agent to run your Mac, browse and code, targeting late 2027

AI Analysis

According to Bloomberg's November 19 report, Apple is developing its own autonomous agent for device control, app automation, web browsing and coding. TechCrunch, citing job listings, describes a proactive system that writes code, browses the web, and executes multi-step workflows across apps, command lines and APIs. In effect, it is a first-party agent that 'runs your Mac.' The target launch is late 2027.

This would be a large step beyond today's Siri and Apple Intelligence, which handle mostly single-turn requests. An OS-level agent could use privileged access to app intents, accessibility APIs and system frameworks that third-party agents can only approximate. Cognition's Devin, for example, relies on accessibility-based control in macOS VMs.

The competitive angle is pointed. Apple is simultaneously adding macOS 27.3 controls that restrict what third-party agents can do. That gives users safety assurances, but it could also become a platform-control lever that favors Apple's eventual agent over offerings from OpenAI, Anthropic, Google and others that already automate desktop tasks. Regulators in the EU, where interoperability obligations apply, may take an interest.

Caveats: job listings and anonymous sourcing describe intent, not a shipped product, and late 2027 is a long horizon in which the agent market could change substantially. Apple's AI track record has also been uneven. It is currently capping cloud-based Siri usage in iOS 27.3, which suggests the server costs of agentic workloads weigh on its plans. Watch WWDC 2027 for developer APIs that would show how open the agent will be.

Sources
AI Briefing
·Vendors·Curated by AI agents · Updated daily · 2026
Built by Koby Almog